White House asks OpenAI to limit GPT-5.6 rollout to government-vetted customers — first preemptive restriction of a major US AI model

The Trump administration has asked OpenAI to restrict the release of GPT-5.6 to a small group of government-approved enterprise customers rather than launching broadly, marking the first time a US administration has preemptively asked an American AI company to limit a model rollout before it goes live.

What's new

The Office of the National Cyber Director and the Office of Science and Technology Policy jointly asked OpenAI to keep GPT-5.6 in a limited preview while government cybersecurity teams evaluate the model. The request came on June 25, following the executive order President Trump signed on June 2 that established a voluntary framework allowing federal teams up to 30 days to assess advanced AI models before general availability.

OpenAI CEO Sam Altman disclosed the arrangement to employees this week, saying the government would be "approving access customer by customer" during a preview period. Altman added that if the limited release goes well, OpenAI hopes to follow with a general, broader release "a couple of weeks later."

According to The Information, OpenAI staffers also "worked closely" with the government on the upcoming release — meaning the arrangement is not purely a restriction but also a collaboration on rollout design.

Context

GPT-5.6 is the model at the center of this restriction. The Trump administration's concern parallels the situation with Anthropic's Claude Mythos 5: that it is capable of identifying and exploiting software vulnerabilities at speeds that no human analyst could match. Mythos was pulled from broad access earlier this month following national security concerns around its advanced cybersecurity capabilities.

The June 2 executive order created the voluntary pre-release review process, giving federal agencies time to test new models against potential misuse scenarios before public release. The GPT-5.6 request is the first formal application of that framework against a specific OpenAI model — and the first instance of the U.S. government proactively asking an American AI company to limit a model's launch before it goes live.

The Trump administration originally positioned itself as taking a hands-off approach to AI regulation. The GPT-5.6 request marks a meaningful departure from that posture, with federal agencies now actively shaping release strategy for frontier models.

Why it matters

For the first time, the federal government has used its voluntary review framework to directly alter how a major AI company stages a product rollout — a practical test of whether the June 2 executive order carries real force. The answer, at least for this launch, appears to be yes.

The "customer by customer" approval process Altman described creates a new dynamic for enterprise buyers: access to frontier models may now depend partly on government clearance. That has implications for how OpenAI — and other frontier labs — structure product launches going forward.

The parallel with Anthropic's Mythos is significant. Together, these two cases suggest Washington is establishing a durable precedent around frontier model releases that extends beyond any single executive order. The question for the industry is whether future administrations maintain this level of involvement — and whether the voluntary framework eventually becomes mandatory.